Regularization of QED by a generalized 't Hooft and Veltman method

Regular izat ion of QED by a generalized 't Hooft and Veltman

method

Marijan Ribaric and Luka Sustersic*
Jozef Stefan Institute, p.p. 3000, 1001 Ljubljana, Slovenia

Abstract

Generalizing the 't Hooft and Veltman method of unitary regulators, we
demonstrate for the first time the existence of local, Lorentz- invariant, phys-
ically motivated Lagrangians of quantum-electrodynamic phenomena such
that: (i) Feynman diagrams are finite and equal the diagrams of QED but
with regularized propagators, (ii) N-point Green functions are causal, (iii) 5-
matrix relates only electrons, positrons and photons, is unitary and Lorentz-
invariant, and conserves charge and total four-momentum.

I. INTRODUCTION

Perturbative predictions about quantum-electrodynamic phenomena implied by a QED
Lagrangian can be computed using the Feynman rules, a regularization method to circum-
vent ultraviolet divergencies, and a renormalization scheme. Regularization method results
in regularized n-point Green functions; a suitable limiting procedure (a renormalization
scheme) then leads to physically sensible predictions that are independent of the particu-
lar regularization method used. But no known regularized n-point Green functions can be
regarded as being based on a physically plausible theory of quantum-electrodynamic phe-

* Corresponding author. Phone +386 61 177 3258; fax +386 61 123 1569; electronic address:
luka. sustersicOijs . si.

1

nomena since the derivation of each disregards some of the basic tenets of conventional
physics (e.g., by lacking a Lagrangian, by not being Lorentz-invariant, by introducing par-
ticles with wrong metric or statistics. . . ). So the perturbative predictions of QED presently
cannot be derived from a physically plausible theory; for a history of, and comments on this
basic problem see, e.g., [|I|. Dirac believed that removal of this conceptual inconsistency
may lead to an important advance in field theories.

To show that one can remove this inconsistency already in four- dimensional space-time,
we will introduce a new, physically motivated modification of the QED Lagrangian and con-
sider it within the theoretical framework of 't Hooft and Veltman : They avoid canonical
formalism and take diagrams as the basis from which everything must be derived; so they
give a perturbative definition of the S-matrix directly in terms of diagrams corresponding
to a given Lagrangian as specified by postulated Feynman rules. The question is: How do
we modify the QED Lagrangian so that the resulting regularized S-matrix is derived from a
theory that is physically plausible?

Gupta 1^ has shown already in 1952 that one can modify the QED Lagrangian so that the
new Lagrangian results in the S-matrix of QED regularized by certain Pauli-Villars method.
And twenty years later 't Hooft and Veltman introduced the method of unitary regulators
(HV-method) that (i) is a variant of Pauli-Villars methods for regularizing propagators,
(ii) requires only an exceedingly simple modification of the initial Lagrangian, and (iii) is
very suitable for proving the causality of the regularized n-point Green functions and the
unitarity of the resulting S-matrix. Unfortunately both methods introduce also unphysical
particles with wrong metric or statistics. To get rid of this serious conceptual deficiency, we
will generalize the HV-method.

We will demonstrate the utility of the generalized HV-method by using it to show that
there is a perturbative S-matrix of quantum-electrodynamic phenomena derived from a
theory that is physically plausible (a pp-theory, for short) in the following sense:

(i) A pp-theory of quantum-electrodynamic phenomena is specified in a continuous, four-

2

dimensional space-time by a local, Lorentz-invariant, physically motivated modifica-
tion of a QED Lagrangian.

(ii) The Feynman rules for this modified Lagrangian, defined as specified by 't Hooft and
Veltman [Q], result in Feynman diagrams that are finite and equal the diagrams of
QED but with regularized propagators.

(iii) All constants of a pp-theory are measurable in principle; there are no auxiliary param-
eters.

(iv) For certain values of these constants, the QED propagators are the same kind of
low-energy approximations to their regularizations as required for renormalization by
Bogoliubov and Shirkov |Q in certain variants of the Pauli-Villars method.

(v) The n-point Green functions of a pp-theory, defined as specified by 't Hooft and Velt-
man in terms of Feynman diagrams, are causal.

(vi) The perturbative S-matrix of a pp-theory, defined as specified by 't Hooft and Veltman
[0] in terms of n-point Green functions, is unitary, Lorentz-invariant, charge and total
four-momentum conserving, and relates only physical electrons, positrons and photons.

Such a pp-theory of quantum-electrodynamic phenomena is not yet known; we cannot in-
corporate a finite-cutoff, Pauli-Villars, dimensional, or lattice regularization of QED in a
pp-theory.

II. LORENTZ-INVARIANT REGULATORS WITHOUT SINGULARITIES

As in the HV-method to each additional singularity of a regularized Feynman propagator
corresponds an additional particle ||^, we will first show how we can Lorentz-invariantly
regularize Feynman propagators so that they do not acquire additional singularities and have
the Kallen-Lehman representation used in proving causality and unitarity Regarding

3

metric and other conventions we follow Refs. in particular, a four- vector k = {k,ik^),
and k^ = k-k- {ky.

Consider a Lorentz-invariantly regularized spin Feynman propagator, say, ^f{x) whose
space-time Fourier transform

(27r)^zA;(A;) = (A;' + - ie)-^ + - ie) (1)

where: (a) the regulator '^{z) is an analytic function of complex variable z with a finite
discontinuity somewhere across the negative real axis; (b) \{z + m?)~^ + ip{z)\ < A\z\~''' for
\z\ > 2m^, with r > 1; (c) ^p{z) is real if 3^2; > and Qz = 0; (d) ip{z) depends on some
constant A so that for any A > Aq > it has properties (a) to (c) with A independent of A,
and, as A ^ cx),

sup — > and sup \(p{z)\ for any 2:0 > . (2)

5ftz>0,S2=0 \z\<zo

As a consequence, the spin propagator provides such a low-energy approximation to its
regularization (|1]) as required for renormalization by Bogoliubov and Shirkov in certain
variants of the Pauli-Villars method.

Using Cauchy's integral formula we can conclude that the Lorentz-invariant regulariza-
tion (|lD of the spin Feynman propagator admits the Kallen-Lehman representation

p-r'^H*)^f (3)

with

p{s) = 6{s - m^) + {2niy^ lini[v9(-s - iy) - (p{-s + iy)] , (4)

s,y > 0. Note that p(s) is real, p(s) = 0{s^^) as s — 00, and

/ sXs) ds = for m = 0, 1, . . . < r - 1 . (5)
Jo

So we can decompose the regularized spin propagator Af{x) into positive and negative
energy parts: Ap{x) = Q{xo)A~^{x) + Q{—Xo)A^{x) 0.

4

The function -i{2'K)-^{y/ - m? + A)"(P + - ie)-^ k'^ + - ie + A)"", A > m,
n = 1, 2, . . ., is an example of a Lorentz-invariantly regularized spin Feynman propagator
that satisfies the above conditions with r = n/2 + 1. Unfortunately, we cannot use such
propagators for a physically plausible regularization of QED since we do not know how to
construct the corresponding local, Lorentz-invariant Lagrangians.

A regulator that satisfies conditions (a)-(c) is by @ a generalization of the Pauli-Villars
regulator that has a continuous mass spectrum. Thus, to use such regulators to construct
a pp-theory, we have to extend the 't Hooft-Veltman construction of Lagrangians in HV-
method to an infinite number of additional fields. To provide an example of how this
can be done, we will present a local, Lorentz-invariant Lagrangian whose propagators for
interacting fields can be taken as spin 1 and spin ^ propagators regularized so that they
acquire no additional singularities and have the Kallen-Lehman representation.

III. LAGRANGIAN THAT REGULARIZES QED PROPAGATORS

Following Veltman , we will consider QED with massive photons in unitary gauge. Its
Lagrangian reads

>Cqed = -\{df,A^ - d^A^y - \^?A^ - tpiYd^, + m)^ + iei^-f^'ijA^ + A^J^ + J^il) + i)Je ,

(6)

where J^{x), Je{x), and Je(x) are four- vector and bispinor source fields, and the non-
vanishing photon mass ^ < 2 x 10"^^ eV 0. The Feynman propagators for the four- vector
field Af^{x) and for the bispinor field ip{x) are:

We could use £qed to define a pp-theory as specified in Section I, were the propagators (|^
faster decreasing when tends to infinity.

However, one can modify the QED Lagrangian so that the propagators for the fields
Afj^ and ip are such regularizations of propagators (|^) that have no additional singularities.

5

Take, for example, the following real-valued, local, Lorentz-invariant Lagrangian:

/:tr = -jC-i - £1/2 + ie^p'y^'^jJA^ + A^J^, + + i>Je (8a)

with

+ qi'si J d'pd'p'f{p'')f{p'W^{x, -p')^'^{x,p)+p'y^,{x, -p')^^{x,p)
-p''^^ix,-p')p"'^,ix,p)], (8b)
A/2 = q^/l J d^p^.^.ix, -p)[At{p^)+p^d^]^,^,{x,p)

- q;,ls^f.J rfVrf'p/(/)/(p')[^i/2(a;, -p')r<il,,,{x,p)p, + c.c] , (8c)

A,ix) ^ I d'pf{p')^,ix,p) , ^{x) ^ J d'pfip')<ily,{x,p) , (8d)

where: \E'^(x,p) and \E'^x,p) are four-vector-valued functions of two four- vectors x and p;
^1/2 (a;, p) is a bispinor- valued function of x and p; 2ad^b = a{dyb) — {d^a)b; ^1/2 = ^1/27^;
t(p^) and f{p^) are real-valued functions of real p^, / d'^pp{p^) = 1; qi, si, gi/2, Si/2, and A
are real constants.

There are three kinds of reasons for the chosen form (H) of the Lagrangian £tr:

(A) It is our purpose to show that there are Lagrangians that generalize the t'Hooft and
Veltman method of unitary regulators 0] to an infinite number of additional fields but
do not introduce additional particles. So we constructed the Lagrangian £tr modify-
ing £qed on the analogy of HV- method 0: (i) We introduced an infinite number of
four- vector and bispinor fields of x that have a continuous index p, namely \E'^(x,p),
\E'^x,p), and \E'i/2(a;,p). (ii) We replaced the free part of Cqed with the free Lagrangian
of these fields, —Ci — C1/2, which is of the first order in d and has a nondiagonal mass
matrix, (iii) In the interaction and source terms of £qed, we replaced the fields A^{x)
and ip{x) with weighted integrals (|55D of \E'^(x,p) and '^i/2ix,p) over the continuous
index p.

6

(B) We tried to simplify the calculations of regularized propagators. We could do without
the four- vector function \l/^a;,p) which we introduced solely to be able to use the
same functions t(p^) and /(p^) in Ci and £1/2- We introduced d to make £tr itself
real- valued, not only its action real as required.

(C) The Euler-Lagrange equations of £tr resemble the Boltzmann integro-differential
transport equation, which can better model rapidly varying, "ultra- high-energy" ,
macroscopic fluid phenomena than the differential equations of motion of fluid dy-
namics. (To this end it uses an infinite number of fields to take some account of
the underlying microscopic behaviour.) Euler-Lagrange equations of £tr may be re-
garded as classical transport equations of motion for the one-particle distribution of
some infinitesimal entities, such as X-ons surmised to underly all physical phenomena
by Feynman ||^. Which all are physical motivations for the type of Lagrangian we
constructed; we discuss them in detail in Ref.

Using the Euler-Lagrange equations of £tr and proceeding as in Ref. [0, we calculate
the dependence of \E'^(x,p) and \E'^x,p) on J^(x), and of \E'i/2(x,p) on Je{x). Thereby we
can infer that the Feynman propagator for the four- vector field defined by ( [8d[ ) equals

9.^^^, (9)

g,{k') = q,s^^hol20 , /^(^') = , (10)

where Imn{k'^) is an analytic function of the complex variable /c^ such that

lUk') = 27:^ y^+^f {y)r^\y)[^ I + A-^k^yt-\y) + l]-^dy (11)
for /c^ > 0; and the Feynman propagator for the bispinor field %l){x) defined by (|8HD equals

-i{27ir%,, L , (12)

^i/2(fc^) = qif2S,,lliol2o . ^k^) = Su\{l - slJliohi + 2k'^lio]}l2o ; (13)

7

where k"^ has to be replaced everywhere with k'^ — ie, by the Feynman prescription.
If functions t(p^) and /(p^) are such that

/ f{y)t{y)\Vy/t{y)r'dy = o (14)

J

for I = 0, —1, . . . , —n, then for complex values of k as oo:

= 0(|fc2|-"/2). (15)

When the function y/t^{y) takes only a finite number of real values Wj, i = 1, 2, . . ., we
can explicitly evalute integrals (pH]) ; we obtain

= ^ + A-2^.A;2 + 1]-- , (16)

where Amni are real constants. Considering such a case, we can show that for any /x^, iv? and
integer n, there exist functions fijP') and t(p^), and constants Si, s^a, 5'i/25 and Ag >
such that the propagators (|^) and with A > Aq are regularizations of spin 1 and spin
I propagators (|^ such that: (i) they have properties analogous to those of propagator (p,
and (ii) there is a positive constant k^ such that for all k'^ > — A^/cq the functions Imn{k'^)i
gi/2{k'^), /^(/c^), giik"^), and m{k'^) are real. In such a case: (i) The constants si, Si/2, Q'l, and
gi/2 are such that

fl^^-ix"^) = ix"^ , m^{-n?)=n?, (17)

g^i^{-ii^) = 1 + djl'^ik'^ = -i/)/dk'^ , gii-m^) = 1 + dm^ik"^ = -m^)/dk'^ . (18)

So the propagators (^ and (p!^ ) have poles at k'^ = —fi"^ and k"^ = —m^, where their
behaviour is given by the spin 1 and spin | propagators (^ with e = 0. (ii) The difference
between spin 1 propagator and propagator (^ depends on the value of A so that it satisfies
relations analogous to (Q); and the same goes for spin | propagators, (iii) The propagators
(P) and ([l^) are analytic functions of fc^ that (a) are not continuous everywhere across
the negative real axis, (b) have no additional singularities to those of spin 1 and spin ^

8

91

o(|A;2|a-")/')

m

9l/2

^2 _j_ ^2

propagators and (c) satisfy relations (|T^). For any integer n > 3, their Kallen-Lehman
integral representations are super convergent: in x-space we can decompose the Feynman
propagators (Q) and ( [T^ ) into positive and negative energy parts without contact terms [Q.

IV. PHYSICALLY PLAUSIBLE REGULARIZATION OF QED

To obtain a perturbative S-matrix of quantum-electrodynamic phenomena based on the
Lagrangian £tr; say ^tr, we use the 't Hooft-Veltman definition of an S-matrix 0]. In view
of results of Sec. Ill, there are functions /(p^) and t(p^), and constants Si, Si/a, qi, qi/2, and
A such that the n-point Green functions of £tr and the corresponding S-matrix S'tr have
the following properties:

(i) As the Lagrangian £tr has the same interaction and source terms as the QED La-
grangian £qed, they are expressed in terms of QED diagrams with the spin 1 and spin
I propagators (0) replaced with their regularizations (||) and (|12|), whereas the vertices
are the same as in QED, i.e., (27r)^7^; so all diagrams are finite.

(ii) As the propagators and (|T^, and higher-order 2-point Green functions of £tr have
no additional singularities, Str relates the same particles as the S-matrix of QED with
massive photons in unitary gauge: electrons and positrons, each with two possible
polarization vectors, and massive photons with three possible polarization vectors;
none of them with wrong metric or statistics.

(iii) N-point Green functions are causal and ^tr is unitary to any order in the fine structure
constant. To show this we may follow the t Hooft-Veltman proof for the HV- method

since the propagators and (|T2D admit the Kallen-Lehman representation.

(iv) Str is charge and total four-momentum conserving since its vertices are such.

(v) S'tr is Lorentz- invariant, since the propagators and (|T2|) commute with Lorentz
transformations and the vertices are Lorentz- invariant.

9

(vi) In the asymptote A — > oo, the propagators (H) and (|^) behave the same way as
required by Bogohubov and Shirkov of the Pauh-Villars-regularized spin 1 and
spin I propagators when their auxihary masses tend to infinity in renormahzation.

So, is a resuh of a pp-theory as defined in Section I. This theory can be regarded as
a physically plausible regularization of QED based on a physically motivated modification
£tr of QED Lagrangian Cq^o-

In view of (vi), we expect that we can compute by renormahzation the renormalized
n-point Green functions of QED with massless photons from the n-point Green functions of
£tr by choosing an appropriate dependence of e, si, Si/2, qi, and gi/a on A, and then limiting
A — s> oo and the renormalized photon mass to zero. However, it is an open question whether
the functions /(p^) and t(p^) must have also some additional properties for the photon third
polarization vector to decouple in the limit of negligible photon mass 0.

V. COMMENTS

Generalizing the 't Hooft and Veltman method of unitary regulators we have shown, for
the first time as far as we know, that one can regularize QED in accordance with the basic
tenets of theoretical physics by suitably modifying the free part of QED Lagrangian. As we
mentioned in Sec. Ill, the physical motivation for such modification has been the Feynman
surmise about X-ons, and the Boltzmann improvement on fiuid dynamics by the transport
theory based on his equation.

Within the framework of quantum field theory as defined by 't Hooft and Veltman 0,
the Lagrangian £tr is related to the physical world solely through the perturbatively defined
scattering matrix ^tr. We see no physical properties of Str that require the spectral function
(^) and the Hamiltonians corresponding to free Lagrangians Ci and C1/2 (which are not free-
particle Lagrangians) to be positive as they turn out to be within the framework of canonical
formalism [^.

The need for a regularization of QED that would result in a physically plausible model

10

was felt very strongly by the founders of QED, Dirac and Heisenberg, already some sixty-
five years ago [0]. But neither they nor their contemporaries succeded in getting rid of
the ultraviolet divergencies by a physicaly motivated modification of the QED Lagrangian.
In the late 1940s, however, Tomonaga, Schwinger and Feynman "solved" the problem of
QED ultraviolet divergencies through renormalization — a solution which does not require
regularization to be physically plausible, and removes all parameters characteristic of regu-
larization. As they obtained spectacularly succesful formulas for quantum-electrodynamic
phenomena, the problem of finding a physically plausible, Lagrangian-based regularization
of QED was not so urgent any more. As there had been no progress whatsoever towards a
solution of this problem, it mainly came to be considered as practically unsolvable [Q] ; those
who hoped otherwise were often considered "irrational", as Isham, Salam, and Strathdee
|[TT| complained twenty-five years later. Thus nowadays, as far as we know, no quantum-field
theorist, excepting the string theorist, pays much attention to this problem, which many of
the preceding generations — e.g., Dirac, Heisenberg, Landau, Pauli, and Salam, to mention
some — still hoped to be solved somehow someday But the string theorists abandon one
of the basic premises of conventional physics, the four- dimensionality of space-time. We
have shown, however, that such drastic steps may be avoided when modifying QED La-
grangian to get rid of ultraviolet divergencies. But the question remains which modification
of the type considered is the most appropriate for better describing quantum-electrodynamic
phenomena at higher energies than the conventional QED, and what is its content.

VI. ACKNOWLEDGEMENT

Authors greatly appreciate discussions with M. Poljsak and B. Bajc and their suggestions.

11

REFERENCES

[1] F. Villars, in M. Fierz and V. F. Weisskopf (Eds.), Theoretical Physics in the Twentieth
Century (Interscience Pubhshers, New York, 1960) p. 78; T. Y. Cao and S. S. Schweber,
Synthese 97 (1993) 33; L. M. Brown (Ed.), Renormahzation (Springer- Verlag, New
York, 1993); S. Weinberg, The Quantum Theory of Fields (Cambridge University Press,
Cambridge, 1995) Vol. I, Sees. 1.3, 7.1, 7.5, and 10.7.

[2] P. A. M. Dirac, in Special Supplement of IAEA Bulletin (IAEA, Vienna, 1969) p.21;
reprinted in A. Salam, Unification of Fundamental Forces (Cambridge University Press,
Cambridge, 1990) p. 125.

[3] G. 't Hooft and M. Veltman, Diagrammar, CERN report 73-9 (1973); reprinted in G.
't Hooft, Under the Spell of Gauge Principle (World Scientific, Singapore, 1994) §2.2.

[4] S. N. Gupta, Proc. Phys. Soc. 66 (1953) 129.

[5] N. N. Bogoliubov and D. V. Shirkov, Introduction to the Theory of Quantized Fields
(Interscience Publishers, New York, 1959) §16.4.

[6] M. Veltman, Diagrammatica (Cambridge University Press, Cambridge, 1994); R.
Lautrup, Nucl. Phys. B105 (1976) 23.

[7] Review of Particle Physics, Eur. Phys. J. C3 (1998) 19.

[8] R. P. Feynman, R. B. Leighton, and M. Sands, The Feynman Lectures on Physics,
Vol. II (Addison- Wesley, Reading, Mass., 1965) Sec. 12-7.

[9] M. Ribaric and L. Sustersic, |hep-th/9810T38| ; Fizika B (Zagreb) 8 (1999) 441

[10] M. Ribaric and L. Sustersic, Fund. Phys. Lett. 7 (1994) 531.

[11] C. J. Isham, A. Salam, and J. Strathdee, Phys. Rev. D5 (1972) 2548.

12